Reply to: 'Potential sources of dataset bias complicate investigation of underdiagnosis by machine learning algorithms' and 'Confounding factors need to be accounted for in assessing bias by machine learning algorithms'.